overview of resulting tools, guidelines, and instruments · in-safety deliverable 3.4 pu contract...

SIXTH FRAMEWORK PROGRAMME PRIORITY 1.6. Sustainable Development, Global Chang e

and Ecosystem 1.6.2: Sustainable Surface Transport

506716

Overview of resulting tools, guidelines, and instruments

Deliverable No. D3.4

Workpackage No. WP3 Workpackage Title New models, tools and guidelines for road safety assessment

Activity No. A3.5 Activity Title Overview of resulting tools, guidelines and instruments

Authors (per company, if more than one company provide it together)

A. Dijkstra, S. Bald, Th. Benz & E. Gaitanidou( editors)

Status (F: final; D: draft; RD: revised draft): F

File Name: IN-SAFETY_Deliverable D3.4.doc

Project start date and duration 01 February 2005, 36 Months

IN-SAFETY Deliverable 3.4 PU Contract N. 506716

February 2008 SWOV

2

Table of contents

List of Figures 3

List of Tables 3

1. Introduction 4 1.1. Simulation models 4 1.2. Risk Analysis Method 5 1.3. Operators’ training schemes 5

2. Relevant methodologies for assessing road safety 6 2.1. Introduction 6 2.2. Crash data, key safety indicators, and crash prediction models 6 2.3. Traffic conflicts 7 2.4. Surrogate safety measures 8 2.5. Comparing design features with design requirements 9 2.6. What if? − analysis 10 2.7. What causes? – analysis 12 2.8. Expert judgement 13 2.9. Selection of methodologies for IN-SAFETY/WP3 14

3. Description of models, methods and tools selected i n this work package 15 3.1. Micro simulation models 15

3.1.1. RutSim 15 3.1.2. S-Paramics 16 3.1.3. VISSIM 17

3.2. Macro simulation models 19 3.2.1. MT.MODEL 19 3.2.2. SATURN 21

3.3. Risk Analysis Method (INRAM) 22 3.4. Operators’ Training schemes 23

4. Guidelines for the application of the models and me thods 25 4.1. Simulation Models 25

4.1.1. Micro Models 25 4.1.2. Macro models 26

4.2. Risk Analysis Method (INRAM) 27 4.3. Operators Training schemes 28

5. Recommendations for further developments 30 5.1. Simulation models 30

5.1.1. RuTSIm 30 5.1.2. S-Paramics 30 5.1.3. VISSIM 31 5.1.4. SATURN 31

5.2. Risk Analysis Tool 32 5.3. Operators’ Training schemes 33

References 34


February 2008 SWOV

3

List of Figures

Figure 1: Components of the Risk Analysis in ADVISORS................................................................... 10 Figure 2: Integration of different disciplines (Durth & Bald 1998) ........................................................ 12 Figure 3: Example for a damage as a result of a concatenation of negative events (Durth & Bald, 1998). 13 Figure 4 System architecture of the traffic simulation tool ......................................................... 18 Figure 5 Two contrary results..................................................................................................... 19 Figure 6 General structure of an assignment model.................................................................. 21 Figure 7: Snapshot of the IN-SAFETY MMT......................................................................................... 29

List of Tables

Table 1: Key indicators for three road types (edited version of Janssen, 2005) ..................................... 7


February 2008 SWOV

4

1. Introduction

Road safety will most probably be influenced by introducing Advanced Driver Assistance Systems (ADAS) or Intelligent Vehicle Safety Systems (IVSS). The effects of these systems on road safety can be assessed in different ways. This document gives a short overview of methodologies which allow for assessing road safety effects (Chapter 2), This Deliverable gives an overview of the outcome of work package 3 of IN-SAFETY. Two methodologies have basically been applied:

• simulation model • risk analysis method

Chapter 3 gives a description of the models and methods, Chapter 4 shows examples of applications of these models and methods.

1.1. Simulation models

Modelling has become a major part of all aspects in traffic within the last decades. The models applied in the IN-SAFETY project range from macroscopic models treating network related facets of traffic to microscopic models which represent traffic flow by moving individual vehicles. The safety aspects can be integrated at several level of modelling targeting different parts of the driver behaviour. The network effects of safety are typically handled by macroscopic models. They represent the supply, the road/street network, and the demand, trips of people, and match both to create the traffic load on the network links. The basic idea of integrating safety is to influence the routing behaviour of drivers in such a way that they observe either the given safety levels on the links of the network or to influence their route choice to minimize the (negative) safety effects of their trips. The first approach adds safety as an additional parameter in the routing/assignment. The information for the safety level comes either from actual traffic data for a specific network or it is derived from known safety indicators for road/street types. The algorithmic extension consists mostly of integrating the supplementary data into the objective function. This approach is, of course, relevant for the drivers, not for other road users. The second approach has a similar basis but tends to optimize individual trips on a network in such a way that the overall safety is maximized. In this nature, the safety of other road users, e.g. pedestrians in a traffic calming zone, are implicitly included. Of course, the safety optimal routes lead not necessarily to a travel time optimal distribution and may thus be different from a system optimum in the traditional sense. Microscopic models are applied for all aspects that directly influence the driving of a vehicle. Generally speaking, driving a vehicle in this context means the driver’s control task in lateral and longitudinal direction. This task can be assisted by new Advanced Driver Assistance Systems (ADAS) or Intelligent Vehicle Safety Systems (IVSS). They take over part of the driving task continuously, like ACC, or only temporarily, like Collision Avoidance, in specific conditions. Such systems lead to changes in the trajectory of the vehicle and thus may


February 2008 SWOV

5

lead to changes in traffic flow as a whole. The analysis of the trajectories in various ways reveal changes in traffic flow as a whole, e.g. changes in the speed-flow-relationship, and also on safety relevant parameters like time-to-collision (TTC) and its derivatives. Other indications like the shape of the headway distribution can also be used for interpretation of safety consequences.

1.2. Risk Analysis Method

The risk analysis methodology applied in IN-SAFETY is based on existing technologies that may estimate safety impacts of new technologies in different road networks. It will help to identify critical situations caused by infrastructure faults or distractive electronic devices in cars. It will help to evaluate positive and negative effects of newly developed driver assistance systems (and their reasons) by prediction of specific risks and shall help road designers and assessors to check more efficiently roads and new technological applications. One main aim of integrating the methodology into IN-SAFETY is to use it with a whole group of researchers, testing how far it allows to enhance the interdisciplinary work. The work in the project was done by the Technical University of Darmstadt by performing the modelling, collecting the data and evaluating the model (in contact with the partners). For analysing the impacts of selected innovations of other work packages, some typical subsystems for the project, e.g. “approaching a sharp bend” (enhancing the available structure) or “changing the lane on a motorway” were selected to be modelled. To do this, information and data are gathered from other partners. All information will be integrated into appropriate models. Data on that/these new model/s is applied. This shall allow to give hints on gaps in safety concerning the system and its implementation. These hints shall permit the other partners to improve their technology and their argumentation for its further use. The methodology was also tested in the work package regarding the pilots, with emphasis on critical road elements, such as sharp bends, tunnels or special highway situations.

1.3. Operators’ training schemes

Within IN-SAFETY research has been performed on operators’ training schemes. Existing schemes, their components and application, as well as their effectiveness have been reviewed through relevant questionnaire throughout the EU. The training of road and TMC operators is a very important procedure so that they would be able to perform their tasks in the most effective level.


February 2008 SWOV

6

2. Relevant methodologies for assessing road safety

2.1. Introduction

Road safety can be assessed in different ways. The most direct way is represented by the crash statistics. It is possible to derive all kinds of risk figures from these crash statistics by combining the number of crashes with the road length or the amount of traffic on a road (type). Crash statistics are only available for existing roads and existing situations. For new roads and for new types of (counter)measures (e.g. ADAS) the safety level or the safety effects can not be assessed by crash figures. Other safety indicators need to be used. One option is to use models which 'predict' the number of crashes given the characteristics of a road (type) or the amount of traffic to be expected. Other safety indicators are based on more indirect measures like the number of conflicts calculated by a microscopic simulation model. A third type of safety indicators is generated by expert knowledge, e.g. a road safety auditor who assesses the safety of a new design by using his experience.

2.2. Crash data, key safety indicators, and crash p rediction models

Crash data Crash data are the most direct way of indicating both the nature of the safety problem and the level of safety. Many tools have been developed for selecting, structuring, analyzing, and visualizing crash data. These tools are useful for existing situations and for existing roads. The nature of crash statistics is that it shows the safety of the past. As soon as one is planning new roads, new types of technical equipment for vehicles, new road facilities, these statistics are of no use anymore. Other indicators are needed for analysing future situations. Key safety indicators Key safety indicators quantify the safety of certain types of roads and junctions. A key safety indicator is determined by relating the absolute level of unsafety (e.g. the number of crashes) on a certain type of road or junction to the degree of exposure. Janssen (1988, 1994) gives a general expression for calculating a key safety indicator:

ExposureofDegree

levelSafetyindicatorsafetyKey =

The safety level is frequently quantified by using crash records. The number of vehicles or the number of vehicle/kilometres is often used to calculate the degree of exposure. An example of a key safety indicator is the number of accidents involving injury per million vehicle kilometres driven. This key safety indicator is also referred to as the risk of a road or junction type. The risk (indicator) based on vehicle kilometres takes into account not only the number of accidents but also the road length and the number of motor vehicles that pass along it (Janssen, 2005).


February 2008 SWOV

7

By combining the length of the road section with the intensity, we can calculate the level of exposure, expressed in millions of vehicle kilometres driven in a year. The level of exposure is then calculated as follows:

4iii 10.65,3*I*LVP −=

in which VPi is the level of exposure of road section i in millions of vehicle kilometres driven in one year, Li is the length of the road section i in km and Ii is the daily volume for road section i. Then, by multiplying the level of exposure VPi by the associated key indicator Ki, the expected number of injury crashes LOi on road section i can be estimated.

iii VP*KLO = The key indicator for road section i depends on the type of road. The key indicators used here for access roads (speed limit 30 kph), distributor roads (50 kph) and through roads within urban areas (70 kph) are shown in Table 1.

Table 1: Key indicators for three road types (edited version of Janssen, 2005)

Road with speed limit in kph Key indicators in number of crashes with injury per billion motor vehicle kilometres

30 122

50 272

70 12

By totalling the calculated, expected injury crashes on the road sections that form part of a route, the total expected injury crashes on the route in question can be derived. Crash Prediction Models Crash Prediction Models are another way of indicating road safety. Using Average Daily Traffic and road characteristics as an input, the number of crashes or casualties can be calculated (FHWA, 2000, 2005; Reurings et al., 2006). The general expression for a crash prediction model is:

,eAADT ijxjii

⋅⋅⋅=γβαµ

where µ is the expected number of crashes in a certain period, AADT is the Annual Average Daily Traffic in the same period, xj are other explanatory variables, α, β, γj are the parameters to be estimated and the subscript i denotes the value of a variable for the i-th road section. Reurings et al. (2006) conclude that (for main roads) the other explanatory variables should at least include the section length, the number of exits, the carriageway width, and the shoulder width.

2.3. Traffic conflicts

Road safety on the level of road sections and junctions is mostly expressed by the number of crashes. However, the number of crashes for a separate road section or junctions is mostly too small for an in-depth analysis to be performed. The number of traffic conflicts and near-


February 2008 SWOV

8

crashes is much higher, therefore enhancing the possibilities for analyses. Studying conflicts and near-crashes presumes a relationship between a conflict and a 'real' crash. This relationship was studied extensively by Hydén (1987) and Svensson (1998). In Germany, Erke et al. (1978) and Zimolong (1982) gave a first overview. Perkins & Harris (1968) started using conflict analysing techniques consequently. Erke et al. �1985� formulated a handbook. The assumption for using this method is that situations with many conflicts have a higher probability for accidents. Trained observers regard traffic situations and analyse and count ´conflicts´. Conflicts are actions of road users, which may lead to problems (e.g. late braking, cutting of bends) and which appear often enough. Several measurements have been proposed to characterize traffic conflicts in detail. For example time to collision (TTC), deceleration rate (DR), encroachment time (ET), post encroachment time (PET), etc. are used to determine the severity of a traffic conflict objectively. �PIARC, 2004� This technique enlarges the amount of data but the used parameters resulting from the manoeuvres are not necessarily direct indicators for risk of accident and reduction of severity.

2.4. Surrogate safety measures

When using a microscopic model conflicts between vehicles will be an integral part of the simulation. The outcome will be used to compare the types of conflicts in a given simulation with the types of conflicts which will be 'acceptable' in a Sustainably-Safe road environment, e.g. conflicts with opposing vehicles should be minimised at high speed differentials. Time To Collision (TTC) is an indicator for the seriousness of a traffic conflict. A traffic conflict is defined by FHWA (2003) as "an observable situation in which two or more road users approach each other in time and space to such an extent that there is a risk of collision if their movements remain unchanged". The TTC value differs between junctions and road sections. A TTC on road sections will only be relevant when one vehicle is following another one. A vehicle on a road section can only have one TTC value. A TTC value on junctions relates to vehicles approaching each other on two different links. A vehicle approaching a junction can have more than one TTC value, depending on the number of vehicles on the other links. Minderhoud and Bovy (2001) have developed two indicators that can be applied in micro simulations and are based on the TTC: the Time Exposed TTC (TExT) and the Time Integrated TTC (TInT). The TExT expresses the duration that the TTC of a vehicle has been below a critical value - TTC* - during a particular period of time. The TExT is thus the sum of the moments that a vehicle has a TTC below the TTC* (figure 1). That means that the smaller the TExT, the shorter time a vehicle is involved in a conflict situation, and therefore, how much safer the traffic situation is. The TExT indicator does not express the extent to which TTC values occur that are lower than the critical value. In order to include the impact of the TTC value, the TInT indicator has been developed. This is the area between TTC* and the TTC that occurs (figure 1). Another way to express the impact of a conflict is to calculate the potential collision energy (PCE) that is released when the vehicles are in conflict and collide with each other. Masses


February 2008 SWOV

9

and speeds of the vehicles, as well as the way in which the vehicles collide with each other, i.e. the conflict type, influence the potential collision energy.

2.5. Comparing design features with design requirem ents

In the Netherlands, the concept 'Sustainably-Safe traffic' (Koornstra et al., 1992; Wegman & Aarts, 2005) is the leading vision in road safety policy and research. The main goal of a Sustainably-Safe road transport system is that only a fraction of the current, annual number of road accident casualties will remain. It is of great importance for a Sustainably-Safe traffic system that, for each of the different road categories, road users know what behaviour is required of them and that they may expect from other road users. Their expectations should be supported by optimising the recognition of the road categories. The three main principles in a Sustainably-Safe traffic system are: • functionality, • homogeneity, • recognition/predictability. The functionality of the traffic system is important to ensure that the actual use of the roads is in accordance with the intended use. This principle led to a road network with only three categories: through roads, distributor roads, and access roads. Each road or street may only have one function; for example, a distributor road may not have any direct dwelling access. The speed limit is an important characteristic of each road category: access roads have low speed limits (30 km/h in urban areas and 60 km/h in rural areas) while through roads have a speed limit of 100 or 120 km/h. The homogeneity is intended to avoid large speed, direction, and mass differences by separating traffic types and, if that is not possible or desirable, by making motorised traffic drive slowly. The third principle is that of the predictability of traffic situations. The design of the road and its environment should promote the recognition, and therefore the predictability, of any possible occurring traffic situations. These principles have been translated into safety design requirements, for instance: For road sections • Avoiding conflicts with oncoming traffic • Avoiding conflicts with crossing traffic • Separating different vehicle types • Avoiding obstacles along the carriageway

For junctions • Avoiding conflicts with crossing traffic • Reducing speed • Limiting the number of different traffic facilities Design requirements in general, which are part of design manuals, are not only based on safety arguments but also on other arguments, like capacity and liveability. In addition a


February 2008 SWOV

10

designer will apply the requirements given the constraints in a real-life situation. Therefore Kooi & Dijkstra (2000) suggested a test to find out the differences between the original safety requirements and the characteristics of the actual design features. This SuSa test systematically compares each deign element or feature with the relevant safety requirements. As a result a percentage shows the total score on Sustainable Safety.

2.6. What if? − analysis

A ´What if? − analysis', or Failure Mode and Effects Analysis (FMEA), aims at revealing possible failures by analysing each component of a system. The method predicts effects of failures for the measure as a whole. The FMEA method was applied in ADVISORS, a research project (2000-2003) co-funded by DG TREN of the European Commission. The main aim of the ADVISORS project was to promote the further development and implementation of Advanced Driver Assistance Systems (ADAS) in Europe. Its aim was to enable the identification of all implementation risks and highlight possible mitigating strategies. A traditional FMEA analysis for technical risk was extended to incorporate other risks like the behavioural, the legal and the organisational risk. The risk analysis of ADVISORS was not very detailed concerning the modelling of the technical context, but embraced a larger scale of other criteria, especially legal and organisational aspects (Figure 1).

Figure 1: Components of the Risk Analysis in ADVISORS

The ADAS function is described and the various risk numbers are determined by different analysis methods. Every risk (technical, behavioural, legal and organisational) can be calculated by the following formula:


February 2008 SWOV

11

Risk Number = Severity x Probability x (Detectability + Recoverability)/2

At the end the Overall Risk Number (ORN) is determined. It is given by the following equation:

32

RNO+RNL+RNB+RNT

=ORN ,

where RNT is Technical Risk Number, RNB is Behavioural Risk Number, RNL is Legal Risk Number, and RNO is Organisational Risk Number Technical Failure Analysis (RNT) Technical assessment considers hardware and software failures, and environmental events. It involves the evaluation of conceivable risks to worker and public safety, and risk of damage to equipment, or environment. Hazards can be identified by using a formal fault and hazard identification process, like the failure mode effect analysis (FMEA). Performing a FMEA starts with defining the system to be analysed, constructing a block diagram and finally identifying all potential items and interface failure modes. Driver Behaviour Analysis (RNB) The analysis method chosen for the driver behaviour analysis is the fault tree analysis followed by the Success Likelihood Index Methodology (SLIM). The underlying idea is that the likelihood of an error occurring is dependent on a relatively small number of Performance Influencing Factors (PIF), like quality of training, procedures and time available for action. It is assumed that experts can judge how good or bad the PIFs are in specific situations. This rating is then multiplied with the relative importance of the PIF and all the outcomes are summed up to create the Success Likelihood Index. This Index predicts the probability of success of the specific situation. Liability and Insurance Analysis (RNL) For assessing the liability and insurance, it is very important to know the legal issues. As it is extremely difficult to obtain “general” legal opinions, it is helpful and necessary to subject specific circumstances to analysis. First, the relevant legal frameworks in terms of traffic and product liability laws and insurance schemes are described. The second step is to identify potential gaps and barriers. The last step is to evaluate and analyse the gaps through interviews, round table discussions and dissemination of preliminary results. So recommendations for legislative action and insurance policies will be formulated and the overall risk can be assessed. Organisational Analysis (RNO) Familiarity and experience in the theory of Work Organisation and Management Structures is required for a detailed analysis. There is no single set of objective criteria for analysing the effectiveness of a particular organisation. To get the needed information about the organisation of the products, in this case ADAS, experts are interviewed by questionnaires during workshops. It is possible to assess the risks and find the severities for several products. [Bekiaris et al., 2001]


February 2008 SWOV

12

2.7. What causes? – analysis

At the Technische Universität Darmstadt risk analysis research for road safety started 20 years ago with two studies (Bald 1991; Durth & Bald 1988). Reason was that conventional methodologies based on accident data could not give enough information to construct prognostic models for the whole process of driving. So primary objective of the work was: • to open up new information sources, especially by making it possible to use data from

the whole process,

• to allow cooperation between different research groups and different disciplines (see Figure 2.2),

• to follow a modular approach to allow continuous improvement without the necessity of reformulating the whole model.

The result was, that in general it is useful to apply the risk analysis for road safety issues. Though it was and is necessary to enhance this approach systematically: The ability to describe human behaviour must be enhanced, e.g. by using non-linear relations.

medicin (psychology), paedagogy

(human)

mechanical engineering

(vehicle)

civil engineering

(road, environment)

ergonomics

road traffic(driving) dynamics(driving) behaviour

medicin (psychology), paedagogy

(human)

mechanical engineering

(vehicle)

civil engineering

(road, environment)

ergonomics

road traffic(driving) dynamics(driving) behaviour

Figure 2: Integration of different disciplines (Durth & Bald 1998)

The described methodology is

• System orientated • Model orientated (allowing to make analysis with parameters beyond the scope of

experimentation) • Modular structured (allowing to exchange, enhance and to detail modules with no or

minimal affection to other modules) • Covering the cause-and-effect-chains allowing to analyse effects of changing parameters The methodology is a systematic extension of the FMEA methodology which allows to describe the non-linear and human behaviour by multidimensional probability distributions. For the use of multidimensional probability distributions, an existing computer program was enhanced to be used on actual computer hardware. With that methodology, it is very easy to deal with large amounts of data (even probability distributions to describe arbitrary behaviour). All available knowledge and data may be used. It is not necessary to throw away any data or knowledge only to make the evaluation process manageable.


February 2008 SWOV

13

The cause-and-effect-chain may be described from the influencing parameters to dangerous situations and accidents. The cause-and-effect-chain takes account of the fact, that most hazards are a result of a concatenation of numerous negative events (see Figure 2.3). It allows to analyse the whole chain and network of relevant facts in single modules, even by different research groups and by different disciplines.

disregard of rulesnonattention etc.

danger prevention isnot possible

accident hasconsequences

area of“active safety”

regular situation

mistake situation

danger stuation

accident situation

damage

disregard of rulesnonattention etc.

danger prevention isnot possible

accident hasconsequences

area of“active safety”

regular situation

mistake situation

danger stuation

accident situation

damage

Figure 3: Example for a damage as a result of a concatenation of negative events (Durth & Bald, 1998).

2.8. Expert judgement

The main objective of safety audits is to ensure that highway schemes operate as safely as possible, i.e. to minimise the number and severity of occurring accidents. This can be achieved by avoiding accident producing elements and by providing suitable accident reducing elements. The purpose of safety audits is to ensure that ‘mistakes’ are not built into new schemes. The UK definition is as follows: A formal procedure for assessing accident potential and safety performance in the provision of new road schemes, and schemes for the improvement and maintenance of existing roads (IHT, 1996). Safety audits can be applied in schemes of various types. Such schemes generally fall into the following categories: • major highway schemes; • minor improvements;


February 2008 SWOV

14

• traffic management schemes; • development schemes; • maintenance works.

The auditor should be a road safety expert, preferably working as part of an audit team. Checklists can be used during an audit. Mostly a set of checklists is available, consisting of specific checklists for each road type.

2.9. Selection of methodologies for IN-SAFETY/WP3

Workpackage 3 of the IN-SAFETY project focussed on a limited number of types of tools for assessing the safety effects of Advanced Driver Assistance Systems: • micro and macro traffic simulation models • Darmstadt Risk Analysis Metthod (INRAM)

Traffic simulation models are very well suitable for using 'Surrogate safety measures'. INRAM is built upon the methodology of 'What causes? – analysis'. Another way of retrieving safety effects from simulation models are 'Crash data, key safety indicators, and crash prediction models'. Crash data can be used for existing roads and streets, given the road users and vehicles of the past. New road sections have no accident history as well as traffic situations with vehicles containing newly developed ADAS. Therefore accident data will used only as a reference for the safety level of existing situations. Key safety indicators can be used for new roads and streets. Thes type of indicators, however, does only take into account one kind of changes caused by ADAS: route choice behaviour (because of changes in traffic flows, which are input for the indicator). Crash prediction models are of the same nature of key safety indicators, the additional value of crash prediction models is that these can be more detailed regarding road and traffic characteristics than crash safety indicators. Nevertheless, crash prediction models do not incorporate characteristics about vehicles with and without ADAS. Finally, Comparing design features with design requirements will be part of the route choice simulation. The physical characteristics of the routes will be compared to the 'ideal' characteristics following form the Sustainable Safety principles. The other methodologies discussed in section 2 will not be applied in this WP. Traffic conflicts require many observations in real-life situations. These kind of observations are not the aim of IN-SAFETY. Expert judgement is mostly used for safety assessment of either new road schemes or new ADAS. This kind of assessment can not easily be integrated in the tools used in this WP. The 'What if? – analysis' was used in the ADVISOR project, the predecessor of IN-SAFETY; see e.g. Bekiaris et al. (2001) for more information about this methodology.


February 2008 SWOV

15

3. Description of models, methods and tools selecte d in this work package

3.1. Micro simulation models

3.1.1. RutSim

RuTSim is a micro-simulation model that consists of sub-models that handle specific tasks. The model is designed to handle one road stretch in each simulation run; i.e., rural road networks are not considered. RuTSim uses a time-based scanning simulation approach. The simulation clock is advanced with a user-defined step size, e.g., 0.1 s. The time-based simulation approach is chosen for RuTSim because it allows more detailed modelling of an individual vehicle’s interactions with the surrounding traffic and the infrastructure. With a shorter time step, the movement of vehicles from one time step to the next becomes smoother and therefore more realistic. Hence, a shorter time step may, given an adequate modelling logic, result in the driving course of events for an individual vehicle to be closer to the driving course of events found in real traffic. The use of a shorter time step does, however, increase the model run time. The model time step should therefore be chosen in relation to the current application. Outputs in the form of aggregated traffic measures do not require a time step as short as that required if a driving course of events for a representative vehicle is desired. The following steps are performed in every time step during a model run: 1. Add vehicles that are to enter the road during the time step to virtual queues, with one

queue for each origin. 2. Load vehicles from the virtual queues to the road, if possible, i.e., if acceptable space

is available on the main road. 3. Update the speed and the position for every vehicle on the road. 4. Remove vehicles that have arrived at their destination. 5. Update the state, i.e., free or car following, overtaking or passed, and acceleration rate,

for every vehicle on the road. 6. Save the data. 7. If animation is enabled, update the graphical user interface (GUI). 8. If the stop time has been reached, terminate the simulation or else increment the

simulation clock and go back to Step 1.

Before the simulation, the speed profile of the road and the traffic that is to enter the road are generated from the input road and traffic data, respectively.

The current version of RuTSim applies a car-following based on the “Intelligent driver model” (Treiber et al., 2000; 2006). This model accounts for driver limitations and anticipation to allow more detailed studies of traffic impacts of driver assistance systems. Overtaking decisions on two-lane highways are controlled by a stochastic model depending on the current road characteristics and the distance to the oncoming vehicle.


February 2008 SWOV

16

3.1.2. S-Paramics

S-Paramics is a microscopic simulation model which is based on a combination of models for car-following, lane-changing, and gap-acceptance. Each individual vehicle is determined by the characteristics of the vehicle (dimensions, speed, acceleration, and deceleration), and by some characteristics of the driver (aggression, awareness). The latter characteristics will influence the overtaking behaviour. Another important element of S-Paramics is the route choice model. The road infrastructure is modelled in a quite detailed way. Nearly each physical element of the road can be modelled. The application of S-Paramics in this work package was focussed on the safety effects of route choice in a road network. Therefore, the more detailed description of S-Paramics refers to the route choice characteristics. In S-Paramics, each vehicle tries to find the shortest route from the road section on which it is located to its destination zone. The shortest route is the one for which the general journey costs are lowest. Each time a vehicle enters a new road section, the route is evaluated again on the basis of the general journey costs that are ‘stored’ in route tables. Familiarity Familiarity with the road network has a fundamental influence on route choice in a hierarchical road network. If this directly influences the quantity of vehicles passing along routes with and without signs, it is important to properly calibrate the level of familiarity. The level of familiarity can be set separately for each vehicle type. For example, if a model includes taxis, it would be quite possible to set the familiarity at 100% because taxi drivers usually know the road network well. General costs The journey costs of an individual road section can be calculated using the general cost comparison. This represents a combination of factors that drivers take into consideration when choosing between various routes. The most important factors are time and distance. If a toll is charged for using certain parts of a road, these costs will also be taken into account. The general journey costs of a road section can be set to the same (generic) value for all vehicles or they can be set by vehicle type. S-Paramics can also calculate the general journey costs for a road category Route tables The route tables are filled in using the general journey costs of the road sections. The route costs are equal to the sum of the general journey costs of the road sections that form part of the route. Route tables give vehicles the opportunity to calculate the costs of a route choice at each junction along the route. When a vehicle approaches a junction, it consults the relevant route table and, after deciding whether to apply perturbation and/or dynamic feedback, the vehicle selects the route that has the lowest journey costs to the destination. Each route table is calculated each time that a simulation is started.


February 2008 SWOV

17

Allocation methods In a one-off operation based on an empty network at the start of a simulation, the all-or-nothing allocation determines the general journey costs for all possible routes associated with a OD relationship. The route with the lowest general journey costs is seen as the shortest route. All vehicles that travel from a given origin to a given destination will then make use of this shortest route. In the stochastic allocation, perturbation is used. Application of this perturbation means that a variance is applied to the general journey costs of a route whenever vehicles have to choose between routes and do not want to make use of an all-or-nothing allocation. In the dynamic feedback allocation, road users who are familiar with the road network take into consideration the congestion in the network when calculating the journey costs of a road section and of routes. Where an all-or-nothing allocation calculates the journey costs on the basis of an empty road network, the dynamic feedback allocation calculates the journey costs of a road section on the basis of the delay imposed by congestion in a constantly revised (dynamic) cost calculation. Road sections that have low journey costs based on calculation for an empty network, and will therefore attract a lot of traffic, will in the course of time produce higher journey costs due to higher concentrations and possibly even congestion. As a result, alternative routes become more attractive. If the congestion decreases, the journey costs of these road sections will decline and become more attractive again. In the dynamic feedback allocation, the various route tables are constantly recalculated for each feedback interval. The stochastic dynamic allocation uses both perturbation and feedback and is therefore the dynamic feedback allocation together with a varying perception of the actual general journey costs (perturbation). It is used in precisely the same way as the stochastic allocation.

3.1.3. VISSIM

VISSIM is able to model transit and traffic flow in urban areas as well as interurban motorways/freeways. The traffic flow model used by VISSIM is a discrete, stochastic, time step based microscopic model, with driver-vehicle-units (DVU) as single entities. The model contains a psycho-physical car following model for longitudinal vehicle movement and a rule-based algorithm for lateral movements (lane changing). Vehicles follow each other in an oscillating process. As a faster vehicle approaches a slower vehicle on a single lane it has to decelerate. The action point of conscious reaction depends on the speed difference, distance and driver dependant behaviour. On multi-lane links vehicles check whether they improve by changing lanes. If so, they check the possibility of finding acceptable gaps on neighbouring lanes. Car following and lane changing together form the traffic flow model. In case of multi-lane roads a hierarchical set of rules is used to model lane changes. A driver has a desire to change lane if he has to drive slower than his desired speed due to a slow leading vehicle or in case of an upcoming junction with a special turning lane. Then the driver checks whether he improves his present situation by changing lanes. Finally he checks whether he can change without generating a dangerous situation. In case of multi-lane


February 2008 SWOV

18

approaches towards intersections this method will lead to evenly used lanes unless routing information forces vehicles to keep lanes. The network geometry is modelled using the graphical user interface of VISSIM. It is possible to load a scanned layout plan of the modelled network or an aerial photo as a background for the network editor. VISSIM models all kinds of different junction layouts and controls like signalized and non-signalized roundabouts and junctions including pedestrians. Vehicle actuated signal control

VISSIM.EXETraffic flow simulation movingcars, trucks, transit andpedestrians through network

VAP.EXEVehicle Actuated Phasingsignal control with *.PUAwhich defines signal groups &phases and *.VAP whichspecifies the control logic

VISVAP.EXEGraphical flow charter

CROSSIG.EXEDesign of fixed timesignal control, phasesand stage transitions

*.PUA *.VAP

Detectorvalues

Signaldisplays

Analysis

- Animation of vehicles- Travel times on segments- Delay + queues at junctions- Cross sectional data- Time-space diagram- Dynamic signal settings

Figure 4 System architecture of the traffic simulation tool

The simulation system itself includes first the traffic flow model and secondly the signal control model. The traffic flow model is the master program which sends second by second detector values to the signal control program. The signal control uses the detector values to decide on the current signal aspects. VISSIM can interface also to several signal controllers which may use the VAP control description language. VISVAP is a graphical designer to set up control logics. CROSSIG is only needed if one has to design very complicated phase transitions. Analysis VISSIM allows a number of analyses both on-line and off-line. For off-line analysis, a number of files is written during the simulation. Most of these files use standard text formats. Some files are written in special formats because of their size and can be processed by analysis tools provided within VISSIM. o Cross Section Measurements Cross section measurements are important for validation of the input values. Especially if profiles (change of input values over time) and routes are used for generating the vehicles, cross section measurements help to check the number of vehicles passing a certain cross section. The cross section measurement file is usually much too detailed for common purposes. VISSIM provides a function to aggregate the raw detector data to a file which includes one


February 2008 SWOV

19

row for each lane location and time interval. The number of vehicles per vehicle type and various data, e.g. speed, within the interval are recorded. o Travel times Travel time is recorded between any two nominated points along a route within the simulation area. The travel time measurement records the average travel time for all vehicles between two points in the network over the recording interval. The travel time is inclusive of all delays including stops at red lights and stops by buses. An examples of travel time graphs is shown for two car movements and two tram movements.

car travel time (500 m)

0

50

100

150

200

250

300

350

400

450

600 1200 1800 2400 3000 3600 4200 4800 5400 6000 6600 7200

simulation time [s]

travel time [s]

fixed SW-NE v.a. SW-NE fixed NW-NE v.a. NW-NE

Figure 5 Two contrary results

3.2. Macro simulation models

3.2.1. MT.MODEL

MT.MODEL is a user friendly and totally integrated system of mathematical models for decision support to the traffic and transport planning. The mathematical models within MT.MODEL offer the opportunity to simulate the variations to the actual mobility and transport planning, previewing the effects that would follow from their realization. The system is composed of: • a data bank, that contains available information of demand and traffic supply; • models of information management; • models of demand and performance prediction of the transport system; • current use and management software. MT.MODEL architecture is based on the general structure of a Decision Support System (DSS). It comprises models for vehicle assignment, multi-user assignment, traffic supply, traffic demand, and OD matrix estimation. The vehicle assignment model, T.ROAD, is composed of: • a traffic supply model for the representation of the road network;


February 2008 SWOV

20

• a traffic demand model for the representation of the desired trips; • models of interaction between demand and traffic supply, simply called “assignment

models”. The traffic supply model allows to represent the road network, according to the oriented graphs theory. The transformation of the graph in network is assured adding a quantitative characteristic of “path cost” to each road section. The traffic supply model, can take into account that the different users categories have similar behaviour, but are characterized by different cost functions. Cost functions are differentiated for users and/or vehicles categories. Every users category has a characteristic cost function. The cost functions of every road section and road category depend on the flows of different categories that pass the section. However, they are independent of any other flow that insists on any other stretch of road. T.ROAD is representing traffic demand by car, using ‘space’ and ‘time’ characteristics. This representation is assured by an OD matrix in a prefixed time period. Assignment models allow to simulate the distribution of vehicular flows on road network. T.ROAD uses a deterministic equilibrium models (D.U.E. – Deterministic User Equilibrium). The deterministic assignment is based on the hypothesis that the users of the road network know perfectly path costs and are rational users. It is demonstrated that the definition on equilibrium flows on the road network can be considered as the solution of a best match problem of an objective function. T.ROAD uses Frank and Wolfe’s algorithm to solve the best match problem. It is an iterative algorithm: to each iteration it produces an executable solution, that is obtained by minimizing a linear approximation of the objective function. The supply model, T.NET, allows in obtaining the fundamental parameters of the table containing the description of the road sections of the network and their features. Therefore the model is a go-between the knowledge of the parameters typical of the road network in the most general form and the descriptive mode of the network required instead by the assignment model. The demand model, T.MOB, can be defined as a mathematical relationship between the average value of the demand in a fixed time period and a certain system of activity and transport. The model gives an estimate of the travel demand in the various time band of the day. The mathematical form of the distribution and modal-split models is a 'multinomial logit'. The OD model, T.OD, estimates OD matrices, using model of mobility demand correction through counts of vehicular or passenger flows. The OD estimation model used by T.OD uses traffic counts on networks and is based on Generalized Least Square method (GLS), which, favours both prior matrix and observed flows. Other inputs can be given to the model as constraints or other objective to be minimised: capacity of links, total of users coming from, or heading to the zones or groups of zones, etc.


February 2008 SWOV

21

3.2.2. SATURN

SATURN (Simulation and Assignment of Traffic to Urban Road Networks) is a traffic assignment model. Traffic assignment models are macroscopic models that simulate traffic by assigning it to different alternative routes (being given an O-D matrix as input) according to several attributes the most important of which being travel time and distance. SATURN offers a wide range of assignment methods including generalised-cost, all-or-nothing, equilibrium, and so on. It employs the main structure of a typical assignment model which is illustrated in Figure 3.3.

Figure 6 General structure of an assignment model

In particular, the work performed within WP 3 involved the influence of the safety level of a route on driver route choice and subsequently in road safety, for urban areas. The tools employed for this investigation were a stated-preference questionnaire, including a stated choice experiment and a traffic assignment model, SATURN. First, within the questionnaire survey drivers had to select the route they would chose for an everyday urban trip, between two alternative routes that were defined through three attributes: travel time, length and safety level. Discrete choice analysis was performed to identify drivers’ preferences. Then the algorithms that described the dynamics of a widely used traffic assignment model were investigated and the way through which the determination of the effect of drivers’ preferences in respect to route choice would alter flow distribution amongst different routes and hence road safety was established. In respect to driver preferences, results revealed that the safety level of a route is not only a contributing factor to route choice, but also the prevalent one, amongst the investigated ones. In addition, drivers’ revealed preferences on the significance of the attributing factors on

TRIP MATRIX

ROAD NETWORK

ROUTE CHOICE

FLOWS, COSTS, etc

ANALYSIS

IN

PU

TS

OU

TP

UT

S

DEMAND SUPPLY


February 2008 SWOV

22

when choosing a route came in accordance with their stated ones, as drivers who choose the safest route rated road safety as the most important parameter. Several demographic characteristics also contribute to drivers choosing the safest route. In particular, drivers tend to choose the safest route more with increasing age and decreasing driving experience. Annual mileage was another contributing factor, as drivers who demonstrated higher exposure would choose more the safe route. Driver education was also found to be a contributing factor, whereas other characteristics such as gender, marital status, household income and profession did not affect drivers’ route choice. The determined significance of road safety reinforces the need for the introduction of road safety into advanced traveller information systems, with the aim of improving road safety. It is expected that drivers being provided with information on the safety level of a route might change their route from less safe to safer ones, which may improve road safety. However, this can be a dynamic equilibrium as the safety level of a route is also dependent on traffic volume, and change in traffic volumes might modify the safety level of a route. To estimate the effects of drivers’ route choice – once safety level is considered as a contributing factor – on road safety, traffic assignment algorithms that consider route safety level need be developed. The results of this study and the aforementioned implications emphasise the need for further research on the contribution of the attribute of road safety on route choice.

3.3. Risk Analysis Method (INRAM)

Road systems are complex processes. Their safety depends on numerous factors (e.g. human behaviour, infrastructure, natural influences, legal factors etc.), which all have to be integrated into the analysis. Especially the need for self-explaining roads requires the integration of human behaviour into the evaluation process. A new methodology should explain the cause-and-effect chains of the complex road system to get logical insight of the ongoing process. To handle the complexity, it should be modular structured to: • allow the integration of the experience of different groups of researchers, also from

different disciplines; • be expandable for all-embracing modelling • be possible to refine for quality enhancement • allow selected improvements without the need for starting from the scratch and should

allow to deal with a huge amount of complex and partly uncertain or quantitative data. In the IN-SAFETY project, TU Darmstadt has further developed its existing methodology and tool and has described it comprehensibly. The aim of the now called Darmstadt Risk Analysis Method (INRAM) is to describe the cause-and-effect chain of critical situations taking into a account the uncertainties of the system (especially human behaviour) allowing to: • identify the reasons for accidents and misbehaviour as well as • test the danger of prospective action alternatives. INRAM is able to analyze complex systems with uncertainty and nonlinear relations. The analysis may be done qualitative, quantitative and in a mixed form. So the methodology


February 2008 SWOV

23

allows the integration of knowledge depending on human behaviour (with subjective measures of value). The created modular models can be refined and expanded. The model and its results are scalable. The same part of the model may be viewed either in a very detailed form or in a more summary overall perspective. The methodology was integrated in the findings of the ADVISORS project coming to an overall covering IN-SAFETY Risk Analysis Method (INRAM) to allow also an assessment of legal and organisational risks.

3.4. Operators’ Training schemes

The training of TMI/TMC and road operators is a very important issue in terms of the proper function of Traffic Management Information Centres and major road networks. Recognising this need, IN-SAFETY tried to identify the current situation in terms of the status of training and the improvements that would be needed to make it more efficient and useful for the operators. Thus, a relevant questionnaire was composed early in the project and sent to relevant parties, in order to collect information on the following topics:

• Current situation of operators training schemes for TMI/TMC, • Description of operators training schemes for TMI/TMC, • Reasons for the Absence of an operators training scheme for TMI/TMC and • Definition of optimal operators training schemes for TMI/TMC.

The results of the responds to this questionnaire showed that in most cases training procedures is that most operators use internal expert training and the training is often on the job, while specific training courses are periodically held. The topics considered in the existing operators training scheme are both normal use and emergency use and contain:

• normal and delay information, • pre-trip information, • emergency information, • basic traffic engineering principles, • incident management procedures, • tunnel safety procedures, • temporary signing principles during maintenance work, • TMS software and hardware training, • electromechanical devices control software training, • maintenance of databases and extensive practice on use of all TMC equipment.

The standards for the performed training are often defined by the key senior staff or the system supplier. It appears that there are no general standards for the training scheme and there is no special authority who is defining the standards.


February 2008 SWOV

24

In most of the questionnaires, the list of the contents considered necessary or useful is much longer than the list of the contents of the existing training schemes. This indicates that many of the operators would be willing to improve their training and they have certain ideas, what should be covered by an improved training procedure. It has also been noted that there is not very much about ADAS/IVIS in the list of existing training contents, however the respondents expressed their interest on various ADAS/IVIS.


February 2008 SWOV

25

4. Guidelines for the application of the models and methods

4.1. Simulation Models

Simulation model in general are meant to evaluate the effects of changes in the traffic and transport system. The changes to be studied will mostly deal with: • growing amount of traffic; • extension of the built-up area; • additional links in the network; • adapting signal control at various junctions. However, in this project the models have been applied for assessing effects of different types of ADAS, given the existing traffic and infrastructure. Micro simulation models can be used for a detailed analysis of (the behaviour of) individual vehicles. Macro simulation models are suited for studying effects on the levels of road sections and road categories.

4.1.1. Micro Models

RutSim The Rural Traffic Simulator, RuTSim, is a traffic simulation model for developed for rural highway environments. The model handles all common types of rural highways including two-lane highways and highways with separated oncoming lanes. The model is designed to handle one road stretch in each simulation run; i.e., rural road networks are not considered. The main road may incorporate intersections and roundabouts, and the traffic on the main road may be interrupted by vehicles entering and leaving the road at intersections located along the simulated stretch. The traffic flows entering the road at various origins may be time dependent. Turn percentages at intersections for each traffic flow are used to determine vehicle destinations. Models in RuTSim are built by experts of VTI in Sweden; RuTSim is not a commercial package. The applications are focussed on research projects. Previous applications of the RuTSim model include quality-of-service studies of alternative rural road designs (Carlsson and Tapani, 2006). RuTSim has also been utilized in a study of possibilities to conduct safety evaluations of driver assistance systems using traffic simulation (Lundgren and Tapani, 2006). S-Paramics and VISSIM


February 2008 SWOV

26

The microscopic simulation models S-Paramics and VISSIM can be used to model (motorized) traffic in urban and rural road networks. The scale of a network in both VISSIM and S-Paramics can vary from one (complicated) junction to a regional network. VISSIM and S-Paramics are used for a range of applications in order to analyse the effects of: • growing traffic demand; • changes in signal control at junctions; • changes in the physical characteristics of the road sections and junctions (adding lanes,

changing speed limits etc.); • adding or removing road sections or junctions; • dynamic route information (in-vehicle and/or on the roadside); • Advanced Driver Assistance Systems (mainly in-vehicle).

A model in VISSIM as well as in S-Paramics will mostly be built and run by a specialist, working at a technical consultancy. Building a model implicates the availability of at least: • detailed road map, preferably digitized; • detailed origin-destination matrix (for a reference year as well as for future years); • traffic counts at strategic cross-sections • data about the signal control at all junctions • speed limit of each road section • observations of queue lengths at junctions and other important stretches of road. Specific for S-Paramics Road safety indicators are no common feature in S-Paramics. However, a microsimulation model produces output regarding position, speed, and direction of each vehicle. These data can be used to calculate time-to-collision values (see section 2.4). The safety indicators can be put together using these values. Specific for VISSIM Also VISSIM has no common features for indicating the level and characteristics of road safety. VISSIM was used to evaluate the effects of different ADAS types. The safety effects were evaluated indirectly by analysing the changes in the average speed: a lower average speed is considered to be positive for road safety.

4.1.2. Macro models

MTModel

MT.Model, a macro simulation model, is a product of CSST. MT.Model is built and run by CSST. MT.MODEL is meant for analyzing the existing situation of the traffic system and, answering to questions as “what if?”, allows to estimate the effects of alternative solutions. In order to obtain a comprehensive evaluation of the effects on the complete transport system, MT.MODEL allows:


February 2008 SWOV

27

• the analysis of the existing situation of demand, traffic supply and performances of the transport system;

• the prediction of the mobility demand with regard to pre-assigned scenarios of socioeconomic and territorial evolution and pre-assigned configuration of traffic and transportation supply;

• the valuation of the performances of transportation networks, according with these scenarios.

Road safety in MT.Model is described by using accident maps, showing the location of accidents on the road map. To estimate the risk factor, CSST uses two methodologies of analysis: • a conventional approach, that is based on the assumption that behaviour of users on

path choice is the same every day: users are sensitive only on travel cost. • an alternative approach. That is based on the assumption that behaviour of users on

path choice can change: use of infrastructure would be improved through an optimized management of traffic flows.

SATURN

SATURN is an existing assignment model which was used by the NTUA. This model is available for researchers and practioners all over the world. There are two functions that are of interest within the framework of IN-SAFETY: the “conventional” assignment model and the network editor. The function of the assignment model assigns traffic, performing trips from an origin to a destination within the simulated road network, to different routes based on a number of principles. The function of the network editor allows and can be applied for the analysis of network-based data which need not be in any way related to traffic assignment problems. Part of the SATURN operation is the estimation of a cost for each route based on which traffic is assigned into different routes. The cost is a function of the travel-time on the route and its distance (length of the route). The additional parameter that needed to be introduced for the incorporation of safety related ITS, is the safety level of the route. Hence, the simulation will now assume that each driver will have knowledge of the time, distance and safety level of each route. Such knowledge is provided through several information systems. Based on this approach the cost function based on which SATURN will assign traffic to the investigated links is given a weight factor for safety.

4.2. Risk Analysis Method (INRAM)

INRAM is FMEA based and allows evaluating complex processes in a modular and interdisciplinary procedure especially by structuring the processes systematically with active and passive elements. INRAM uses Numerically Described Multidimensional Probability Distributions (NDMPD) to describe data and provides a Database of Knowledge (DoKn).


February 2008 SWOV

28

For making NDMPDs tradable, a tool called Darmstadt Risk Analysis Tool (INRAT) is provided. INRAT is principally not limited to a certain number of dimensions and elements and so restricted only by available computer memory and calculation time, allowing the model to evolve as needed. To investigate the safety of the whole traffic system it must be divided into appropriate subsystems, e.g. straight road (different cross-sections), sharp bends, and intersections. How to work with this methodology? • Identify a subsystem you are interested in or which is relevant to be analysed. • Try to describe all the facts and relations you know (especially concerning driver

behaviour) with a network of active and passive elements. Integrate other disciplines and research groups where appropriate.

• Try to describe the situations and developments in this network with the available knowledge. If you assume relations, but have no exact data, try to use realistic assumptions (it is better to make assumptions, rather than not taking these relations into account at all; the description by multidimensional distributions allows to consider numerous relations, especially human behaviour, it also allows to do additional calculations for different assumptions without significantly more expense).

• Work with that model. Test it. Refine it. • Use the model for practical purposes. One main advantage of risk analysis is to start with a more general description of the road traffic system and gradually refine it. All assumptions may be replaced by results of advanced knowledge of research. Therefore it is possible to start with simple descriptions of the real road traffic system.

4.3. Operators Training schemes

The IN-SAFETY project has developed a complementary operators training curriculum, consisting of the “Operators’ Training Manual” and the “Operators’ Training Multimedia Tool - MMT”. Both items are aiming to provide additional information mainly on ITS, their applications and impacts, in order to assist operators, but also higher personnel in their operative and decision making processes. More specifically, the “Operators Training Manual” can be considered as an additional textbook to be provided to operators during their training, which would provide them with useful knowledge on innovative technological applications in the field of road safety, their identified impacts, as well as relevant standards. Moreover, background knowledge on transport engineering is also provided.


February 2008 SWOV

29

Figure 7: Snapshot of the IN-SAFETY MMT

Additionally, the Manual, together with the MMT can be very useful consulting tools in decision making processes, i.e. when considering the installation of some new equipment to the infrastructure, or when this has already been installed, at the first stages of application, so as to get information on its functionality, possible drawbacks or special requirements. This would be of use not only for operators but also for the decision makers in a TMI/TMC or a road operation centre. By this work, it is considered that a significant amount of knowledge on ITS and other road safety related new technologies is being communicated and the operators are provided with a useful consulting as well as training set of tools.


February 2008 SWOV

30

5. Recommendations for further developments

5.1. Simulation models

5.1.1. RuTSIm

The rural highway traffic simulation model RuTSim has been extended to allow simulation studies of driver assistance systems. The extensions made to the model are introduction of variable overtaking behaviour, a possibility to specify different categories of drivers/vehicles and inclusion of time-to-collision based safety indicators among the standard model output. In the extended RuTSim model, different driver/vehicle categories can be used to create traffic including different driver types and/or vehicles equipped with different driver assistance systems. An indication of the road safety effect of the systems is then given by the safety indicators included among the model output. The implementation of these model extensions is verified through a simulation study of effects of an in-vehicle overtaking assistant system. Future research using the RuTSim model includes analysis of traffic effects of the individual driver results of the VTI IN-SAFETY pilot. These results include changes in driving performance due to driver fatigue and rumble strips on two-lane highways. In addition, further research on the relation between simulation based safety indicators and accident risks is needed in order to facilitate more accurate safety analysis using traffic simulation models.

5.1.2. S-Paramics

Route choice is a complicated part of travelling, and it is not simple to model this behaviour. This study has only dealt with the road safety effects of route choice. These effects can be determined in different ways: in this report a number of conflict measures and the route diagram were developed and applied. Other indicators are still being studied in current research. The results of the different indicators do not all point in the same direction. The significance of the different indicators for research into route choice requires more attention. What is more, it is important to examine whether the indicators in the micro simulation provide a result that conforms to reality. The indicators for the safety of a route only comprise the safety of car drivers using a route. The indicators will be extended to the safety of all users (also cyclists and pedestrians) of a route. Making a route safer does not necessarily mean that the whole network will become safer: unsafety can move to other routes. That is why a method will be developed for optimizing the safety of all (main) routes in the network.


February 2008 SWOV

31

For planning applications the method should be integrated in existing planning models. For traffic management applications the safety criteria should be built into the choice algorithms of route planners.

5.1.3. VISSIM

A microscopic model like VISSIM is a sound basis for determining any phenomena in traffic flow. Applications with more sophisticated requirements on the level of detail of modelling are those about safety aspects. This is due to several facts, the two most prominent being the small number of safety critical conditions and the demands on the precision of the vehicles’ movements. Unfortunately, these two requirements are to some extend contradictory: while the first one implies large sample sizes with corresponding long run times, the second one implies more detail in the modelling of the vehicle and driver behaviour requiring more time for each interaction. Further development could therefore lead into two directions: • Creating safety indicators from a less detailed model for a large sample size (in space

and/or time) by applying additional information on the criticality of certain situations • Further detailing of available simulation approaches or even development of new

driver-vehicle models with special regard to critical situations. To this end some basic research into the driver behaviour in special stress situation sis required. Such research involves considerable efforts, especially due to the reproducibility of results.

It seems possible to combine both approaches: a cruder model is applied to create “general” traffic flow and a more detailed model takes over in certain conditions which could lead to safety critical situation. An analogous approach had been tried in the past to combine macroscopic and microscopic traffic models. However, with the ever increasing computing power, such approaches with their particular problems have been really useful only for the time period until enough computing power was available to allow micro-models to do the job alone. So the major focus should be on increasing the level of detail for modelling the driver behaviour in safety critical conditions. An aspect of particular importance is the fact that averages do not matter in terms of safety: it is the last few percentages of distributions that are important – whether in parameters describing driver behaviour or in the probabilities for conditions to appear. This leads to large efforts for creating driver models for critical situations with realistic parameter sets. However, it seems worthwhile to tackle this because simulation is the only possibility to investigate safety aspects without endangering human lives.

5.1.4. SATURN

One limitation of this application that needs to be considered is the fact that the DATA input is fixed and independent of the flow levels. One may argue though that the safety levels of a route are dependent on the respective flow levels of the route. Hence, a mechanism has to be employed to deal with this shortcoming of the proposed application.


February 2008 SWOV

32

To apply SATURN in the most efficient way the DATA input, hence in the discussed application the road safety level of the simulated routes, is fixed and independent of traffic flows. However, it is argued that the traffic flow on a link affects the road safety level of the link. This means that the initial flow attributed to each route based on the cost that is estimated for each route (taking into account travel-time, distance and safety level), will affect not only the travel-time on the link but also its safety level. In the SATURN simulation, the second assignment iteration will recalculate the travel-times of each route but not the safety level, as this is fixed. To overcome this shortcoming, the code of SATURN could be modified to allow for the DATA input to be linked with the assigned traffic flows and to be modified according to a user defined formula following each assignment iteration. For this, a formula that relates road safety and traffic flows needs to be used.

5.2. Risk Analysis Tool

It is necessary to analyse the long-term effect of new infrastructure, regulations and accessories with all-embracing risk analysis methods which are able to integrate the effects of human behaviour and habits. As new endangerment may arise after the first safety successes have become apparent. E.g. regarding ADAS systems: if such systems are useful and effective but not reliable, new risk may arise if the user is trusting the system but the system fails and the has no chance to remark the failure in time. It seems useful to build an overall covering model of the road system as most of the behavioural aspects are cross-linked throughout the system. The modelling process may be started at different points, letting the different parts gradually grow together. The model may temporarily branch if reliable knowledge is not yet available within certain sections. But always, the goal should be to integrate all road related knowledge into one model (and its adjacent Database of Knowledge). Such a model could be used from all disciplines to enable and simplify the process of problem analysis, discussion of variants and assessment of political recommendations. • The goal for the future should be to add additional modelled scenarios (e.g. tunnel

sections), to expand the scope of the modelled scenarios and to refine them increasingly. Increased attention should be given to international and interdisciplinary work.

• To promote the international and interdisciplinary work, it is necessary to establish a community. Center of this community is a Database of Knowledge (DoKn), which has to be established, e.g. as a worldwide spanned distributed database system, and a system of common procedures, to interact between the researchers and this database, e.g. like the Request-for-Comments (RfC) procedure of the internet. The process may be additionally promoted by mailing lists, wikis and other interactive tools.

• The calculating tool should be enhanced to be more comfortable to the users (input format, integration into the Database of Knowledge) and to be more powerful (faster, e.g. by using computer clusters; larger datasets etc.)


February 2008 SWOV

33

5.3. Operators’ Training schemes

As for the Operators’ Training schemes, it can be suggested that tools as the ones developed within IN-SAFETY could be integrated in the operators’ training existing or under development curricula, for further information on the new technological achievements in the field of ITS and other road safety and efficiency related systems. These tools should of course be regularly updated, so as to include state-of-the-art information on the included technologies and/or any relevant updates and follow ups. Thus, the operating centres and their personnel would be able to use a reliable tool that would enhance their decision making procedures towards introducing new technological equipment to the infrastructure, selecting the appropriate one for their needs and also dealing with the in-vehicle technological innovation which have subsequent influence on the safe and efficient function of the road networks under their responsibility. The overall aim is to make the road networks as forgiving and self-explaining as possible, showing to their operators the technological alternatives available for this purpose.


February 2008 SWOV

34

References

Bald, J. (1991). Grundlagen für die Anwendung von Risikoanalysen im Straßenwesen. Dissertation. Darmstadt. Bekiaris, E. et al. (2001). Compendium of existing Insurance schemes and Laws, risk analysis of ADA systems and expected driver behavioural changes. User awareness enhancement, dissemination report and market Analysis and ADAS marketing strategy ADVISORS. D3/8.1 V4. Aristotle University of Technology, Thessaloniki. Carlsson, A. & Tapani, A. (2006), Rural Highway Design through Traffic Simulation, In: Proceedings of the 5th International Symposium of Highway Capacity and Quality of Service, Yokohama, Japan. Durth, W. & Bald, J. (1988). Risikoanalysen im Straßenwesen. Darmstadt 1987. Forschung Straßenbau und Straßenverkehrstechnik Heft 531. BMVBW, Bonn. Erke, H. et.al. (1978). Verkehrskonflikttechniken im Innerortsbereich. Unfall- und Sicherheitsforschung Straßenverkehr Nr. 15. Bundesanstalt für Straβenwesen, Bonn. Erke, H. et.al. (1985). Verkehrskonflikttechnik – Handbuch für die Durchführung und Auswertung von Erhebungen. in Unfall- und Sicherheitsforschung Straßenverkehr. Nr. 52. Bundesanstalt für Straβenwesen, Bonn. FHWA (2000). Prediction of the expected safety performance of rural two-lane highways. FHWA-RD-99-207, McLean, Virginia, USA. FHWA (2003), Surrogate Safety Measures From Traffic Simulation Models. Report nr. FHWA-RD-03-050, McLean, Virginia, USA. FHWA (2005). Validation of accident models for intersections. FHWA-RD-03-037, McLean, Virginia, USA. Hydén, C. (1987). The development of a method for traffic safety evaluation. Lund University, Lund. IHT. (1996). Guidelines for the Safety Audit of Highway. The Institution of Highways and Transportation, London. Janssen, S.T.M.C. (1988). De verkeersonveiligheid van wegtypen in 1986 en 2010 (Road safety related to road types in 1986 and 2010). R-88-3. SWOV, Leidschendam.


February 2008 SWOV

35

Janssen, S.T.M.C. (1994). Safety standards based on road type. In: Proceedings of the 3rd International Conference on Safety and the Environment in the 21st Century: lessons from the past, shaping the future, Tel Aviv, November 7-10, 1994, pp 79-88. Janssen, S.T.M.C. (2005). De verkeersveiligheidsverkenner gebruikt in de regio; de rekenmethode en de aannamen daarin (The road safety explorer used regionally; the calculation method and its assumptions). R-2005-6. SWOV Institute for Road Safety Research, Leidschendam. Koornstra, M.J., Mathijssen, M.P.M., Mulder, J.A.G., Roszbach, R. & Wegman, F.C.M. (1992). Naar een duurzaam veilig wegverkeer (Towards a sustainably safe traffic system). Nationale verkeersveiligheidsverkenning voor de jaren 1990/2010. SWOV, Leidschendam. Minderhoud, M.M. & Bovy, P.H.L (2001). Extended time-to-collision measures for road traffic safety assessment. In: Accident Analysis and Prevention, Volume 33, pp. 89-97. PIARC (2004). Road Safety Manual. World Road Association PIARC, Paris. Perkins, S.R. & Harris, J.L. (1968). Traffic conflict characteristics – Accident potential at intersections. Highway Research Record 225. Transportation Research Board. Washington, DC Reurings, M., Janssen, T., Eenink, R., Elvik, R., Cardoso, J. & Stefan, C. (2006). Accident prediction models and road safety impact assessment; a state-of-the-art. First deliverable of WP2 of the Ripcord-Iserest project. Svensson, Å. (1998). A method for analysing the traffic process in a safety perspective. Lund University, Lund. Treiber, M., Hennecke, A. & Helbing, D. (2000). Congested traffic states in empirical observations and microscopic simulations. Physical Review E 62(2), pp. 1805–1824. Treiber, M., Kesting, A. & Helbing, D. (2006). Delays, inaccuracies and anticipation in microscopic traffic models. Physica A: Statistical Mechanics and its Applications 360, pp. 71–88. Wegman, F. & Aarts, L. (eds.) (2006). Advancing Sustainable Safety. National Road Safety Outlook 2005 - 2020. SWOV, Leidschendam. Zimolong, B. (1982). Verkehrskonflikttechnik – Grundlagen und Anwendungsbeispiele. Unfall- und Sicherheitsforschung Straßenverkehr. Nr. 35. Bundesanstalt für Straβenwesen, Bonn.

overview of resulting tools, guidelines, and instruments · in-safety deliverable 3.4 pu contract...

Documents